Research Design
The methodological approach that was utilized for this study is a non-
experimental, cross-sectional research method. Non-experimental research can be either descriptive or correlational. A descriptive study can describe and organize variables or characteristics of the sample without examining the relationships between the variables or attempt to infer/predict the variables under investigation in order to provide information on behaviors, attitudes, or other characteristics of a particular group (Salkind, 2013). Correlational research measures the degree (strength and nature) of the relationship between two variables (Salkind, 2013).
Correlation coefficients are generalities, i.e., it describes the entire group, not every individual in the sample. Correlational designs do not imply causation but focus on describing relationships between variables. Since correlational designs explore the strength of the relationship between two continuous variables, it is able to indicate whether the relationship is positive (as one variable increases, so does the other) or negative (as one variable increases, the other decreases). Furthermore, other statistical tests that will be employed in data analysis for this study are Logistic Regressions, two- way, between-groups analysis of variances (two-way ANOVAs), and Multiple
Regression.
Logistic Regression is used to predict an outcome that is categorical from predictor variables that are continuous and/or categorical (Salkind, 2013). For instance, comparing different conditions (e.g. social support, loneliness, depressive symptoms) on a variety of outcome measures (e.g. drug/alcohol use, sexual behaviors, HIV infection
status). The two-way ANOVA design compares the mean differences between groups that have been split on two independent variables. A two-way ANOVA allows
researchers to understand if there is the possibility of an interaction effect between the two independent variables on the dependent variable (Salkind, 2013). For instance, this researcher can understand if there is an interaction between partner status and religiosity on levels of social support amongst middle-aged HMSM; hence, partner status (partnered or not) and religiosity (religious affiliation or not) are the independent variables and social support is the dependent variable. Multiple regression allows the researcher to learn more about the relationship between several independent variables and a dependent variable (Salkind, 2013). This allows the researcher to evaluate the importance of each of the independent variables in the model and test the overall fit of the model with the data collected.
Independent or Predictor Variables
The independent or predictor variable refers to the intervention or treatment that may be manipulated (Salkind, 2013). The independent variable can be evaluated with respect as to how it is correlated and compared to the dependent variable. In this study, the independent variables were loneliness, depressive symptoms, substance use/abuse, and social support.
Dependent or Outcome Variables
The dependent or outcome variable is the main characteristic being measured in the study (Salkind, 2013). The dependent variable is the principal variable of interest in the study and may be affected by the independent variable. In this study, the dependent variable was sexual behaviors.
Attribute Variables
Attribute or demographic variables are pre-existing characteristics that describe the sample in the study. The attribute variables in this study included categorical and continuous variables. The continuous data included age, years of residence in the US, offspring (if any), and number of years of education. Categorical data included relationship status, living arrangements, nationality, self-identified sexual orientation, religion, employment status, income, health insurance status, and perceived health status.
Sample
As demographics change, the US has become a rapidly growing multicultural population. Hispanics account for more than half of the growth in the total population between 2000 and 2010, making them the nation's largest ethnic minority (CDC, 2011). According to the US Census Bureau (2011), it is estimated that 52 million Hispanics reside in the US, representing approximately 16.7% of the US total population. It is projected that the Hispanic population will reach 132.8 million by 2050 (30% of the US population).
Setting
South Florida has a large population of Hispanics (US Census, 2011). Since South Florida has a large gay community, the chosen setting was Miami-Dade County. The Florida Department of Health in Miami-Dade County (2010) estimates that 1 in 97 Hispanic men in Miami-Dade County is living with HIV infection. The July 2009 census reported that Miami-Dade County included a population of 2,500,625 of which 48.3% are male and 61.4% are Hispanic; furthermore, reported male non-family households
this study, Hispanic gay and bisexual men were grouped into the category of HMSM. According to Lieb et al. (2006), there are 1,455,490 Hispanic men in Florida of which 4 to 10% are MSM. Therefore, the HMSM population of Florida is estimated to be between 58,220 to 145,549.
Increased recruitment of middle age HMSM participants was noted after
identifying venues that middle age HMSM frequent and placing flyers in each location. Following the Institutional Review Board (IRB) approval from Florida International University, flyers promoting the study were posted at multiple sites across South Florida including organizations, bars, barbershops, immigration offices, local stores and street fairs, where middle age HMSM are known to congregate. There are numerous gay bars in Miami-Dade County that operate seven days a week with ranging working hours. Prior to placing flyers and utilizing these establishments and venues, this researcher obtained verbal consent from the owners or managers of these establishments.
Selection of Participants
To accurately represent the population of interest in this study, the selection of participants is vital. Because it is impossible to locate and survey the entire middle age HMSM population of Miami-Dade county, a representative subset with a certain number of participants from the middle age HMSM population was obtained as the sample for this study. Additionally, a non-probability sampling technique of convenience sampling was utilized.
According to Burns and Grove (2009), convenience sampling is considered a weak sampling method, since it is not random; however, convenience sampling is the most common approach used in nursing research studies. Using convenience sampling
does not ensure that the derived sample is representative of the larger population. Despite the weaknesses of convenience sampling, it is less expensive, less time consuming, and useful in this proposed study, whereby the researcher can readily access a sample of middle age HMSM.
The sample size is calculated after the study’s significance level, effect size, and power are determined. The alpha or significance level is typically set at 0.05. Alpha is the probability of rejecting a true null hypothesis and making a Type I error. Type I errors infer reporting that there is a finding when indeed there is not. Since the alpha for this study was set at 0.05, there is a five percent chance of committing a Type I error (Polit & Beck, 2012).
The study’s power refers to the probability of rejecting the null hypothesis, thereby avoiding a Type II error. Type II errors infer reporting that there is not a finding when indeed there is a finding. Since a power of 80% is generally considered acceptable for research studies, this was selected for this study (Polit & Beck, 2012).
Effect size is a proposed measure that explains the magnitude of the difference between groups, but does not take into account the variability in scores (Sullivan & Feinn, 2012). According to Cohen (1987), effect sizes are classified as small (d = 0.2), medium (d = 0.5), and large (d ≥ 0.8), in which a medium effect of 0.5 is visible to a vigilant observer. Even if the statistical analysis is significant, if the two study groups' means do not differ by at least 0.2 standard deviations, the difference is minor (Sullivan & Feinn, 2012).
The researcher may estimate the effect size if they are unsure of the associations of the variables under investigation (Cohen, 1987). Since effect size is the main finding
of a quantitative study and this study employed Multiple Regression as its most powerful statistical method, a medium effect size of 0.5 was chosen (MacCallum, Browne, & Cai, 2006). However, the researcher realized that this designation does not take into account the accuracy of the instruments and the diversity of the study population, instead provides a general guide. Lastly, according to desired significance level of 0.05 and the power of 0.80, the sample size should be no less than 150 participants for an estimated effect size of 0.5 (moderate difference effect) (Polit & Beck, 2012). Since there resulted a wide range of possible responses, some variables were collapsed and converted to dichotomous variables; however, this will be discussed in Chapter 4.
Inclusion Criteria
Inclusion criteria help the researcher define and identify the characteristics needed to answer the research questions. Inclusion criteria for participation in this study
included: (a) self-identification as a Hispanic or Latino man; (b) self-identification as gay/homosexual or bisexual; (c) 40 to 65 years of age (middle aged); (d) required to be able to listen or read and comprehend English or Spanish; and (e) currently reside in South Florida before enrolling in the study.
Exclusion Criteria
Exclusion criteria help the researcher identify characteristics of the population that are not desired in the study. Participants may have certain conditions that may limit participation. Exclusion criteria in this study included participants who: (a) refuse to self- identify as Hispanic or Latino men; (b) do not self-identify as MSM; (c) are not 40 to 65 years of age; (d) cannot comprehend English or Spanish; and (e) do not reside in Miami- Dade county.
Instrumentation
Data on middle aged HMSM’s attitudes and behaviors were collected using five instruments. The following are the instruments that were used in this study:
Multidimensional Scale of Perceived Social Support (MSPSS), the University of
California, Los Angeles (UCLA) Loneliness Scale, the Behavioral Risk Assessment Tool (BRAT), the Center for Epidemiologic Studies Depression (CES-D) Scale, and a standard demographic questionnaire. The paper packet with the five instruments was collated into a booklet entitled “Hispanic Men’s Health Study.”
Demographic Questionnaire
A demographic questionnaire was used to collect both categorical and continuous data. The continuous data included age, years of residence in the US, number of
offspring (if any), and number of years of education. Categorical data included relationship status, living arrangements, nationality, self-identified sexual orientation, religion, employment status, income, health insurance status, and perceived health status. A faculty member from the University of Miami School of Nursing & Health Studies translated the demographic questions professionally and it was made available to this researcher for use in this study.
Multidimensional Scale of Perceived Support (MSPSS) Scale
The variable of social support was measured using the MSPSS instrument. The MSPSS instrument is designed to measure perceptions of support for an individual from three sources: family, friends, and a significant other (Zimet et al., 1988). The MSPSS scale was designed as a research instrument that incorporates ease of use and participant time conservation.
Description of the MSPSS Scale
The MSPSS scale is a 12-item instrument with four items for each subscale (family, friends, and significant other). The items that Zimet et al. (1988) included in the MSPSS scale were developed from research literature on social support, which identified an inverse relationship of social support with anxiety and depression. Since the MSPSS scale is easy to use, conserves time, and has a strong foundation in research, it has been widely used as the instrument to measure social support in research studies where social support is a variable under investigation. The MSPSS scale can be completed in
approximately 5-minutes and requires approximately 2-minutes to score. Since the MSPSS is in the public domain in both English and Spanish versions, it does not require permission for use.
Norms of the MSPSS Scale
The MSPSS scale was normed with 275 undergraduate university students (mean age of 18.6, SD = .88) of which 136 were female and 139 were male. From this 275 undergraduate sample, 69 participated in test-retest procedures during instrument development. Additionally, Liu and Mustanski (2012) utilized the MSPSS instrument with 248 lesbian, gay, bisexual and transgender (LGBT) youth (47.7% male) with a mean age of 18.76 years (SD = 1.34) and with 11.4% Hispanic participants. The MSPSS demonstrated excellent internal consistency for family (Cronbach’s α = .90) and peer support (Cronbach’s α = .91).
Scoring of the MSPSS Scale
The MSPSS scale can be scored for each subscale individually or used as the total scale score. The total score measures the overall support of an individual. There are four
items in each subscale (family, friends, and significant other) that have to be tallied then divided by four to result in a subscale score. The total scale score is calculated by adding all of the items then dividing them by 12. Items are rated on a 7-point Likert-scale ranging from 1 (very strongly disagree) to 7 (very strongly agree). Zimet et al. (1988) suggest that a mean scale score ranging from 1 to 2.9 could be considered low support; a score of 3 to 5 could be considered moderate support; and a score from 5.1 to 7 could be considered high support.
Reliability Data on the MSPSS Scale
Numerous studies have shown the MSPSS to have good internal and test-retest reliability, good validity, and a stable factorial structure (Mustanski & Lui, 2012). The MSPSS is a summation instrument with Cronbach’s alphas ranging from .85 to .91 for each subscale and the reliability of the total scale was .88 (Zimet et al., 1988). High internal consistency was found with test-retest reliability over 2- to 3-month interval ranging from .72 to .85 for the subscales, and .85 reliability for the total scale (Zimet et al., 1988).
Validity of the MSPSS Scale
The MSPSS scale was found to have excellent concurrent validity. Since Zimet et al. (1988) hypothesized that a lack of perceived social support would relate to anxiety and depression, the subscales for depression and anxiety in the Hopkins Symptoms Checklist (HSCL) were used to construct validity. Zimet et al. (1988) reported correlations
between the MSPSS scale and the HSCL supported the following prediction, which constructed validity: perceived support from family was significantly inversely related to depression (r = -.24, p < .01) and anxiety (r = -.18, p < .01); perceived support from
friends was significantly inversely related only to depression (r = -.24, p < .01); perceived support from a significant other was significantly inversely related only to depression (r = -.13, p < .05); and the summated perceived total support scale was significantly inversely related only to depression (r = -.25, p < .01).
Spanish Version of the MSPSS Scale
According to Marin and Marin (1991), research with bilingual Hispanic
populations has shown that participants prefer to respond to questionnaires in Spanish. Therefore, the Spanish version of the instrument was utilized. Fortunately, the MSPSS scale was translated into Spanish and is available in the public domain (American Academy of Pediatrics, 2010).
University of California, Los Angeles (UCLA) Loneliness Scale
The variable of loneliness was measured using the UCLA Loneliness scale, version three. The UCLA Loneliness scale is designed to measure self-reports of concurrent loneliness as well as feelings of social isolation (Russell et al., 1978). The UCLA Loneliness scale was designed as a research instrument that has been widely used with MSM, including HMSM (Grov, Golub, Parsons, Brennan, & Karpiak, 2010; Martin & Knox, 1997; Sandfort et al., 2007).
Description of the UCLA Loneliness Scale
The UCLA Loneliness scale is a 20-item instrument measuring an individual’s subjective feelings of loneliness and social isolation. Russell et al. (1978) developed the items on the scale from an in-depth review of other loneliness scales and research literature on both loneliness and social isolation. Currently, the UCLA Loneliness scale is in its third version, which is the instrument that was utilized in this study. The UCLA
Loneliness scale is easy to use and has a strong foundation in research. It has been widely used as the instrument to measure loneliness in research studies where loneliness is a variable under investigation (Grov et al., 2010; Martin & Knox, 1997; Sandfort et al., 2007). The UCLA Loneliness scale can be completed in approximately 8-minutes and requires approximately 3-minutes to score. Since the UCLA Loneliness scale is in the public domain in both English and Spanish versions, it does not require permission for use.
Norms of the UCLA Loneliness Scale
As part of a larger study investigating loneliness, the UCLA Loneliness scale was normed with 492 university students, 227 from UCLA (76 males and 151 females) and 265 from Tulsa University (130 males and 135 females). One-hundred and two student volunteers from Tulsa University participated in a two month test-retest procedure during instrument development. In a multi-city research study with self-identified HMSM, Sandfort et al. (2007) reported an internal consistency of 0.78 for the adapted UCLA Loneliness scale with 302 HMSM from Miami, 309 HMSM from New York and 301 HMSM from California. Additionally, Hubach et al. (2015) studied factors of loneliness, HIV infection related stigma, and condom use in a sample of 100 HIV infected MSM residing in rural Indiana. The variable of loneliness was assessed using the revised third version of the UCLA Loneliness scale. Composite measures for loneliness using the UCLA Loneliness scale resulted in Cronbach’s α = 0.941 (Hubach et al., 2015).
Scoring of the UCLA Loneliness Scale
The UCLA Loneliness scale contains 20 items, which is scored using a 4-point letter scale. The following are the ranges for the 4-point letter scale responses: O
signifies “I often feel this way”, S signifies “I sometimes feel this way”), R signifies “I
rarely feel this way, and N signifies “I never feel this way.” Instructions indicate to score
all of the letters O and N as zero, R as one and S as two, suggesting that higher total scale scores indicate lower self ratings of satisfaction and happiness with increased loneliness (Russell et al., 1978).
Reliability Data on the UCLA Loneliness Scale
Numerous studies have shown the UCLA Loneliness scale to have good internal and test-retest reliability, good validity, and a stable factorial structure (Russell et al., 1978; Russell, Peplau, & Cutrona, 1980; Russell, 1980; Weeks, Michela, Peplau, & Bragg, 1980). The UCLA Loneliness scale is a summation instrument with a Cronbach alpha of .96 for the first version and ranged from .89 to .94 in its third version (Russell et al., 1978; Russell et al., 1980). The third version of the UCLA Loneliness scale was tested using 301 elderly participants (over 65 years of age), of which 121 were male and 180 were females. With this elderly sample, the UCLA Loneliness scale was re-
administered 12 months later and resulted with a high internal consistency and test-retest reliability correlations scores equal to .73, which signified a similar finding found with prior psychometrical testing (Russell et al., 1980).
Validity of the UCLA Loneliness Scale
The UCLA Loneliness scale was found to have concurrent validity. Several validity criteria were used to examine the UCLA Loneliness scale. First, Russell et al. (1978) compared results among groups (clinic versus student scores), which resulted (r (45) = .79, p < .001) for the self-report question about current loneliness and the loneliness scale. Since loneliness has been linked to other emotional states, such as
depression and anxiety, the researchers significantly correlated their findings with the Beck Depression scale (1967) and its measures of depression (r (131) = .49, p < .001) and anxiety (r (131) = .35, p < .001). Lastly, Russell et al. (1978) significantly correlated their findings with the anxiety subscale from the Multiple Affect Adjective Checklist by Zuckerman and Lubin (1965) resulting in (r (65) = 43, p < .01). Therefore, the UCLA Loneliness scale supports its theoretical framework by linking loneliness to emotional states such as depression, anxiety, feelings of boredom and emptiness.
Spanish Version of the UCLA Loneliness Scale
Research employing the Spanish version of the UCLA Loneliness scale has reported good reliability and validity with a stable factor analysis. For instance, Borges, Prieto, Ricchetti, Hernandez-Jorge, and Rodriguez-Naveiras (2008) administered the Spanish version of the UCLA Loneliness scale to 522 undergraduate university Spanish and Italian students, whose median age was 19.6, and reported that the scale had a
theoretical and analytical coherent bifactorial structure with good reliability and validity. Therefore, the Spanish version of the UCLA Loneliness scale was utilized in this study. The UCLA Loneliness scale is available in the public domain in Spanish (Borges et al., 2008).
Behavioral Risk Assessment Tool (BRAT)
Developed collaborative from a CDC funded five state and 5-year project (Prevention With HIV-Infected Persons Project [PHIPP]), the BRAT was designed to collect HIV risk behaviors, including sexual and drug/alcohol behaviors, homelessness, incarceration, and HIV testing history (Wisconsin HIV Prevention Evaluation Work
Group, 2000). Both sexual and drug/alcohol related behaviors were measured using the BRAT. In this study, this instrument was the most informational and time-consuming.
Description of the BRAT
The BRAT is a 61-item instrument with categorical and a few continuous
variables, i.e., number of sexual partners by gender; years with main partner if applicable; and the number of times and how many drug, alcohol and/or tobacco used in the past 30